3 Clinical evidence

3 Clinical evidence

Summary of clinical evidence

3.1 Full details of all clinical outcomes considered by the Committee are available in the assessment report overview.

3.2 The key clinical outcomes for WatchBP Home A presented in the decision problem were:

  • diagnostic accuracy for hypertension

  • diagnostic accuracy for atrial fibrillation

  • incidence of atrial fibrillation-related stroke in people in whom atrial fibrillation is detected

  • reduced mortality from atrial fibrillation-related stroke

  • reduced disability from atrial fibrillation-related stroke

  • device-related adverse events.

3.3 The clinical evidence for WatchBP Home A was based on 5 studies focusing mainly on the diagnostic accuracy of the WatchBP Home A algorithm to detect atrial fibrillation, compared with the recognised standard of 12-lead ECG. These included 3 cross-sectional diagnostic studies, a small case series (19 patients) with diagnostic outcomes, and an unpublished comparative diagnostic cohort study. An ongoing comparative diagnostic study was identified as relevant to the decision problem but no results were available for consideration.

3.4 The External Assessment Centre used data from 2 further studies evaluating the diagnostic accuracy of manual pulse palpation for atrial fibrillation against 12-lead ECG (Cooke et al., 2006; Hobbs et al., 2005) to make a comparison with the WatchBP Home A device.

3.5 The study by Wiesel et al. (2004) compared the sensitivity, specificity and diagnostic accuracy of a home blood pressure monitoring device (the Omron 712C automatic sphygmomanometer), modified to detect pulse irregularity (and hence atrial fibrillation), with 12-lead ECG. The study used the same algorithm as that included in the WatchBP Home A device. The study included an initial cohort of 125 inpatients from a hospital in New York, USA, and a main cohort of 450 outpatients from a US cardiology clinic, all of whom had had 12-lead ECGs before the intervention. Of these patients, 53 inpatients and 54 outpatients had atrial fibrillation diagnosed on ECG examination. Patients with a pacemaker were excluded from the study, as were patients for whom data were not available from 2 readings. The included outpatients (n=446) had 2 successive readings taken with the modified sphygmomanometer during a scheduled clinic visit. An irregularity index, defined as the standard deviation of the time interval between beats divided by the mean of the time interval, was used to analyse ECGs from the inpatient cohort. A threshold irregularity index was selected as a value that all ECGs showing atrial fibrillation would exceed. The threshold index was then used to determine the diagnostic accuracy of the device; the rhythm was considered to be irregular if 2 successive, paired readings were greater than the threshold index. The External Assessment Centre noted that the protocol to detect atrial fibrillation in this study was defined as 2 successive positive results, as opposed to the 'usual' mode of the WatchBP Home A, which is defined as 3 successive positive readings. An analysis of 446 paired readings showed that the device had a sensitivity of 100%, a specificity of 91%, and a diagnostic accuracy of 92% for detecting atrial fibrillation. The confidence intervals and statistical significance of these results were not reported.

3.6 Wiesel et al. (2009) conducted a study on 405 cardiology outpatients visiting 2 clinics in the US, to determine the diagnostic accuracy of a Microlife oscillometric automatic home blood pressure monitor (BP3MQ1). This device used the same algorithm as that included in WatchBP Home A, to detect pulse irregularity likely to be atrial fibrillation. The study population was considered representative of people at risk of atrial fibrillation. A 12-lead ECG and 3 sequential readings with the home monitor were taken for each patient during their visit to the clinic. Two positives out of 3 readings was considered indicative of atrial fibrillation. Patients with pacemakers or defibrillators were excluded. The irregularity index and threshold value identified in the earlier study by Wiesel et al. (2004) were used to determine diagnostic accuracy (see section 3.5). Of the 405 patients tested, 93 (23%) patients had atrial fibrillation diagnosed by ECG readings and of these, 90 were correctly identified using the home monitor after 3 readings. For single readings, the home monitor had a sensitivity of 95% (95% confidence interval [CI] 93 to 98) and a specificity of 86% (95% CI 84 to 89). For 3 sequential readings, it had a sensitivity of 97% (95% CI 91 to 99) and a specificity of 89% (95% CI 85 to 92). Of the 64 patients with abnormal rhythms not attributable to atrial fibrillation on the ECG, the home monitor correctly classified them as non-atrial fibrillation in over 50% of cases overall. The specificity of the device for patients in sinus rhythm was 97%.

3.7 Stergiou et al. (2009) performed a study to determine the diagnostic accuracy of a Microlife oscillometric automatic home blood pressure monitor (BPA100 Plus) including the same algorithm as that in WatchBP Home A to detect atrial fibrillation. The study population was recruited from people attending an outpatient hypertension clinic or admitted to a university medical ward in Athens, Greece, and also included healthy volunteers. In total, 73 patients were recruited, including 27 with known persistent atrial fibrillation, 23 with non-atrial fibrillation arrhythmias, and 23 with sinus rhythm to act as controls. Patients with pacemakers or defibrillators were excluded. For each patient, 3 successive blood pressure readings were taken using the home monitor, while a 12-lead ECG was recorded simultaneously during each measurement. The algorithm used the irregularity index and threshold value identified in the study by Wiesel et al. (2004) (see section 3.5). The sensitivity and specificity of the home monitor for atrial fibrillation diagnosis were assessed for single, duplicate and triplicate measurements. A total of 217 simultaneous blood pressure measurements and ECG recordings were obtained from 73 patients. The sensitivity of the device was 93% (95% CI 74 to 99) and the specificity was 89% (95% CI 76 to 96) for detecting atrial fibrillation from a single measurement. If 1 positive reading from 2 taken was needed to indicate atrial fibrillation, the sensitivity was 100% (95% CI 84 to100) and specificity was 76% (95% CI 60 to 87). For triplicate measurements, if 2 positive readings were needed to detect atrial fibrillation, the sensitivity was 100% (95% CI 84 to 100) and specificity was 89% (95% CI 75 to 96). The authors concluded that 2 positive readings out of 3 taken was the optimal diagnostic mode for the device. Of 5 false positives observed, all showed some heartbeat irregularity during ECG measurement.

3.8 Wiesel et al. (2007) carried out a case series study to determine the efficacy of a modified home blood pressure monitor (Omron 712C automatic sphygmomanometer), with the same embedded algorithm as that included in WatchBP Home A, to detect atrial fibrillation when used by patients at home over an extended period of up to 5 months. A group of 19 patients, who were in sinus rhythm at their clinic visit, but with at least 1 previous episode of atrial fibrillation, was recruited from a hospital-based clinic in the US. The mean patient age was 74. During the study period, 6 patients were receiving antiarrhythmic medication and 11 were receiving warfarin. Patients were given the device to monitor their blood pressure once per day for up to 30 days if no irregularity was detected. If irregularity was detected, patients were asked to repeat their readings up to 3 times and in the case of 3 positive readings, to return to the clinic for an ECG. Patients were monitored for between 5 days and 5 months and an ECG was carried out at each clinic visit regardless of the home monitor readings. The monitor correctly identified recurrent atrial fibrillation in 7 patients. Throughout the study period, 9 patients had no irregular readings and in 3 patients irregular readings were identified because of sinus arrhythmia or ectopy; these were therefore false positives. One patient with atrial fibrillation had intermittently regular readings, which may have been atrial flutter.

3.9 Wiesel et al. carried out a comparative study in the USA. The Committee considered detailed findings from this study in an unpublished paper presented as academic-in-confidence (Wiesel et al., 2012). The findings have been presented as a poster (Wiesel, Saji and Messineo, 2010). The authors compared the accuracy of a Microlife home blood pressure monitor containing the WatchBP Home A algorithm with an ECG event monitor (Heartrak 2) in diagnosing atrial fibrillation in a home setting. Patients (n=160) aged 65 or over, or with hypertension, diabetes, congestive heart failure or a history of stroke, were given both devices to use at home for 30 days. The patients were instructed to follow a protocol in which home monitor and ECG readings were recorded and transmitted to the investigators. ECG recordings were used to determine the diagnostic accuracy of the home monitor. After a series of positive monitor readings, a second ECG was taken and used to confirm a true or false positive reading. Analysis was based on 139 patients, after considering withdrawals, exclusions and incomplete data. Of these, 16 patients had a history of previous atrial fibrillation. A total of 3896 home monitor readings were taken with corresponding ECG readings. The home monitor detected atrial fibrillation with a sensitivity of 98.5% (95% CI 92.8 to 99.9) and a specificity of 91.7% (95% CI 91.4 to 92.3). The device detected atrial fibrillation in all patients who had at least 1 atrial fibrillation-positive ECG, and in 8 patients with a history of atrial fibrillation who were diagnosed with the condition during the study period and who followed the protocol correctly. The device also detected atrial fibrillation in 2 asymptomatic patients with no history of atrial fibrillation, whose diagnosis was confirmed by ECG. The home monitor also made 7 false positive diagnoses.

3.10 An unpublished report containing interim data from a pilot study carried out in 15 GP practices in the north Hull locality group, covering a population of 54,000 (19% of the total primary care trust population) was also considered (Hancocks, M: personal communication, 2012). The GP practices used 80 WatchBP Home A devices over a 6 month period. Atrial fibrillation prevalence was compared between practices using the WatchBP Home A monitors and other localities with a broadly similar profile. Overall, 160 new cases of atrial fibrillation were identified across all Hull practices, 71 (44%) of which were made using WatchBP Home A. This equates to an increase in atrial fibrillation prevalence from 1.17% to 1.22% across all practices. Prevalence in the practices using WatchBP Home A increased by 0.8% compared with 0.4% in those not using the device.

Committee considerations

3.11 The Committee considered that the studies described above, although conducted in a hospital setting, provided evidence that the WatchBP Home A device was able to detect pulse irregularity in people having their blood pressure measured. The Committee then considered the setting in which the device would be used and concluded that the most likely benefit would be from using the WatchBP Home A device in primary care to detect asymptomatic atrial fibrillation during blood pressure measurement, rather than as a home monitoring device, for the reasons described in 3.12–3.15.

3.12 The Committee noted that ambulatory blood pressure monitoring is the preferred method for establishing a diagnosis of hypertension in people with raised blood pressure measured in the clinic, in line with the NICE clinical guideline 127 on hypertension. It considered that there was not enough evidence to justify using the WatchBP Home A device in place of ambulatory blood pressure monitoring for suspected hypertension.

3.13 The Committee noted that evidence that the WatchBP Home A device could detect atrial fibrillation in people having their blood pressure measured at home was limited to a small case series and an unpublished study.

3.14 The Committee concluded that because the NICE clinical guideline on hypertension recommends that home blood pressure monitoring should be undertaken twice daily for between 4 and 7 days, it would be unlikely that the device would detect asymptomatic, paroxysmal atrial fibrillation under these circumstances.

3.15 The Committee considered that the most likely benefit from using the device in primary care would be to increase the rate of detection of atrial fibrillation in people having their blood pressure measured and thereby reducing stroke incidence. Although the Committee acknowledged that there were limited data to support the clinical utility of this scenario, it was persuaded by data from the pilot study in GP practices in north Hull. This study showed that using the WatchBP Home A device may as much as double the detection rate of atrial fibrillation when used in this setting, compared with manual pulse palpation.

3.16 The Committee noted that in order to access the treatment pathway recommended in NICE clinical guideline 36 on atrial fibrillation, people with suspected atrial fibrillation detected by the WatchBP Home A device should have an ECG.

  • National Institute for Health and Care Excellence (NICE)